45
© 2013 IBM Corporation 1 Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor [email protected] IBM Research – Haifa

Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor [email protected] IBM Research

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation1

Surviving the Worst: A Vision for OpenStack Disaster Recovery

Presenter: Michael Factor

[email protected]

IBM Research – Haifa

Page 2: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation2

Outline and Presentation Objective

Basic Disaster Recovery (DR) concepts

• What is Disaster Recovery (DR)?

• Recovery time objective (RTO), recovery point objective (RPO)

• Data replication

• Consistency

Workload Example

Vision for OpenStack DR

Objective: Motivate the requirements for DR for OpenStack and encourage involvement in on-

going efforts to enable OpenStack DR.

Page 3: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation3

What is Disaster Recovery?

According to Wikipedia, Disaster Recovery (DR) is “the process, policies and procedures . . . for recovery . . . of technology infrastructure . . . after a natural or human-induced disaster.”

Page 4: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation4

What is Disaster Recovery?

According to Wikipedia, Disaster Recovery (DR) is "the process, policies and procedures . . . for recovery . . . of technology infrastructure . . . after a natural or

human-induced disaster.”

Implication: Surviving a disaster requires geographic dispersion

Page 5: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation5

What is Disaster Recovery?

According to Wikipedia, Disaster Recovery (DR) is "the process, policies and procedures . . . for recovery . . . of technology infrastructure . . . after a natural or human-induced disaster.”

ConfigurationSoftwareNetworkStorageServers

Page 6: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation6

What is Disaster Recovery?

According to Wikipedia, Disaster Recovery (DR) is "the process, policies and procedures . . . for recovery . . . of technology infrastructure . . . after a natural or human-induced disaster.”

Up front (good path)

• Planning

• Copy

• Testing

DR Copy

Periodic

Test

Primary DC Secondary DC

Page 7: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation7

What is Disaster Recovery?

According to Wikipedia, Disaster Recovery (DR) is "the process, policies and procedures . . . for recovery . . . of technology infrastructure . . . after a natural or human-induced disaster.”

Up front (good path)

• Planning

• Copy

• Testing

DR Copy

Periodic

Test

X

Primary DC Secondary DC

Copy Approaches• Continuous

• Synchronous

• Asynchronous• Periodic

• Online• Offline

Page 8: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation8

What is Disaster Recovery?

According to Wikipedia, Disaster Recovery (DR) is "the process, policies and procedures . . . for recovery . . . of technology infrastructure . . . after a natural or human-induced disaster.”

Up front (good path)

• Planning

• Copy

• Testing

Detection

DR Copy

Periodic

Test

X

Oops!

Primary DC Secondary DC

Page 9: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation9

What is Disaster Recovery?

According to Wikipedia, Disaster Recovery (DR) is "the process, policies and procedures . . . for recovery . . . of technology infrastructure . . . after a natural or human-induced disaster.”

Up front (good path)

• Planning

• Copy

• Testing

Detection

Recovery

• Infrastructure

• Application

Secondary DC

Page 10: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation10

Recovery Point Objective (RPO)Recovery Time Objective (RTO)

Recovery Point Objective Recovery Time Objective

How far back in time a

disaster takes one

How long until operational

after a disaster

RPO=0 � synchronous copy RTO=0 � hot backup site

Page 11: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation11

Replicating the data: Synchronous

A

ACK

A

Primary Secondary

1 4

2

3

Page 12: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation12

Replicating the data: Asynchronous

A

ACK

A

Primary Secondary

1 2

3

4

Page 13: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation13

Consistency

Consistency: “Data at secondary is what the application could have seen at the primary”

• May require some “fix-up” to make consistent

If application writes A,B,C,D,E

• Having A,B,C at secondary is consistent

• Having E,B,D at secondary is inconsistent

Inconsistent data results when the data is not forwarded in the order of hosts writes

• Primarily an issue with asynchronous replication

Inconsistent data is essentially garbage

Page 14: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation14

Outline

Basic Disaster Recovery (DR) concepts

Workload Example

• A three tier application

• What is needed for the workload to survive a disaster?

Vision for OpenStack DR

Page 15: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation15

Workload example: Three Tier Application

Client

Persistentstorage

Client Client

AppServer

AppServer

AppServer

Host

DataServer

Host

AppServer

AppServer

AppServer

Host

Data

Server

Host

Persistentstorage

Image

Repo

Identity

Service

Page 16: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation16

Client

Persistentstorage

Client Client

Host

Host

Host

Host

Persistentstorage

AppServer

AppServer

AppServer

AppServer

AppServer

AppServer

DataServer

Data

Server

• Ensure image content is available at recovery data

center.• Ensure compatible image

metadata

What is needed for the workload to survive a disaster?

Image

Repo

Images for Application and Data Servers obtained from Glance

Identity

Service

Page 17: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation17

Image

Repo

Application and Data servers

VMs managed by Nova

Security managed by Keystone

Network configured by NeutronIdentity

Service

Client

Persistentstorage

Client Client

Persistentstorage

What is needed for the workload to survive a disaster?

AppServer

AppServer

AppServer

Host

DataServer

Host

AppServer

AppServer

AppServer

Host

Data

Server

Host

• Compatible metadata for network,

security and VMs at the recovery site

• Consistent with application’s persistent data

• May be different approaches for

different metadata

Page 18: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation18

ClientClient Client

Host

Host

Host

Host

AppServer

AppServer

AppServer

AppServer

AppServer

AppServer

DataServer

Data

Server

What is needed for the workload to survive a disaster?

Persistentstorage

Persistentstorage

Image

Repo

Persistent storage managed by Cinder

Identity

Service• Persistent state modified by

application should be replicated to secondary

• Configuration information needs to be replicated, e.g.,

volume size

Page 19: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation19

ClientClient Client

Host

Host

Host

Host

AppServer

AppServer

AppServer

AppServer

AppServer

AppServer

DataServer

Data

Server

What is needed for the workload to survive a disaster?

Persistentstorage

Persistentstorage

Image

Repo

Heat can be a means of extracting configuration information and deploying at recovery site

Templates from primary and recovery site may not be identical

Identity

Service

Page 20: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation20

Outline

Basic Disaster Recovery (DR) concepts

Workload Example

Vision for OpenStack DR

• Overview

• State

Images

Data

Metadata

• Automate

Page 21: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation21

Some basic tenants of our vision

The DR is between a primary cloud and a target cloud

• Independent of one another – share-nothing

Except perhaps a geo-distributed Swift deployment

• Primary and target clouds interact through a “mediator”.

Enable hybrid deployments between private and public cloud

• Also private-private and public-public

Protect a set of VMs and related resources

• A workload or a set of workloads owned by a tenant.

Allow flexibility in choice of RPO and RTO

Page 22: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation22

Vision for OpenStack DR: The big picture

Security

Storage

VMs

Images

Network Security

Storage

VMs

Images

Network

DR Middleware DR MiddlewareWorkload

description

Workload

descriptionSecurity details, VM images, and

metadata, etc (Icehouse and beyond)

Primary cloud Target cloud

Storage StorageStorage Replication (Icehouse)

Page 23: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation23

State: Images Glance registry

• Handle like metadata

Glance backend store

• Swift global cluster

• Cinder

Storage replication

• Manual replication

Page 24: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation24

State: Images

Proxy Layer

Storage Servers

Private

Network

Swift Global Cluster

Glance registry

• Handle like metadata

Glance backend store

• Swift global cluster

• Cinder

Storage replication

• Manual replication

Page 25: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation25

State: Images

Proxy Layer

Storage Servers

Private

Network

Swift Global Cluster

Glance registry

• Handle like metadata

Glance backend store

• Swift global cluster

• Cinder

Storage replication

• Manual replication

Page 26: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation26

State: Images

Proxy Layer

Storage Servers

Private

Network

Swift Global Cluster

Glance registry

• Handle like metadata

Glance backend store

• Swift global cluster

• Cinder

Storage replication

• Manual replication

Page 27: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation27

State: Data High RPO

• Back up to Swift

Global/Remote Cluster

Low RPO

• Storage level replication

Cinder managed (storage)

Nova managed (host)

Page 28: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation28

State: Data High RPO

• Back up to Swift

Global/Remote Cluster

Low RPO

• Storage level replication

Cinder managed (storage)

Nova managed (host)Cinder

Volume

Repl.

Gateway

Driver

Cinder

VolumeDriver

I need storage

replicated in

Europe

Page 29: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation29

State: Data High RPO

• Back up to Swift

Global/Remote Cluster

Low RPO

• Storage level replication

Cinder managed (storage)

Nova managed (host)Cinder

Volume

Repl.

Gateway

Driver

Cinder

VolumeDriver

Scheduler locates driver

that supports appropriate local-remote pairing

Driver creates local copy

and performs any initialization for pairing

Page 30: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation30

State: Data High RPO

• Back up to Swift

Global/Remote Cluster

Low RPO

• Storage level replication

Cinder managed (storage)

Nova managed (host)Cinder

Volume

Repl.

Gateway

Driver

Cinder

VolumeDriver

Cinder asks Gateway to

create remote volume via remote Cinder

Page 31: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation31

State: Data High RPO

• Back up to Swift Global

Cluster

Low RPO

• Storage level replication

Cinder managed (storage)

Nova managed (host)Cinder

Volume

Repl.

Gateway

Driver

Cinder

VolumeDriver

All host writes are now replicated

Work for Icehouse

Page 32: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation32

State: Metadata

Examples of OpenStack metadata

Nova

• VM flavors

• SSH keys

Keystone

• Identities of users

Neutron

• Virtual networks between VMs

Cinder

• Volume types

• Pairing

Glance Registry

• Image metadata

Approaches to replicating metadata

• Periodic

• Continuous

Transfer and apply at remote site

• Some “clean up” required

• Some commands applied at once others only at recovery

Copying raw data from controller DBs does

not work

• Non-selective

• Requires same configuration/hardware at both sites

Page 33: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation33

Automation

Identify what to protect and setup

• ImagesEnsure content available at secondary

• MetadataExtract description and dependencies

Replicate

Create dependencies

Set up to deploy

• DataSet up for replication

Test

_________________________

Failover

• Deploy at secondary

Page 34: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation34

Secondary

Glance

Automation Example: Glance

Proxy Layer

Storage Servers

Private Network

A3B5

Swift Global Cluster

Primary

Glance

1. Create image in Swift Global

Cluster

Page 35: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation35

Secondary

Glance

Automation Example: Glance

Proxy Layer

Storage Servers

Private Network

Swift Global Cluster

A3B5

Primary

Glance

1. Create image in Swift Global

Cluster

Page 36: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation36

Secondary

Glance

Automation Example: Glance

Proxy Layer

Storage Servers

Private Network

Swift Global Cluster

A3B5

Primary

Glance

1. Create image in Swift Global

Cluster

A3B5

Page 37: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation37

Secondary

Glance

Automation Example: Glance

Proxy Layer

Storage Servers

Private Network

Swift Global Cluster

Primary

Glance

1. Create image in Swift Global

Cluster2. Define primary Glance to point to

image in Swift

A3B5

A3B5

A3B5

Page 38: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation38

Secondary

Glance

Automation Example: Glance

Proxy Layer

Storage Servers

Private Network

Swift Global Cluster

Primary

Glance

1. Create image in Swift Global

Cluster2. Define primary Glance to point to

image in Swift

A3B5

A3B5

A3B5

Page 39: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation39

Secondary

Glance

Automation Example: Glance

Proxy Layer

Storage Servers

Private Network

Swift Global Cluster

Primary

Glance

1. Create image in Swift Global

Cluster2. Define primary Glance to point to

image in Swift3. Extract metadata from primary

A3B5

A3B5

A3B5

A3B5

Page 40: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation40

Secondary

Glance

Automation Example: Glance

Proxy Layer

Storage Servers

Private Network

Swift Global Cluster

Primary

Glance

1. Create image in Swift Global

Cluster2. Define primary Glance to point to

image in Swift3. Extract metadata from primary

4. Replicate to secondary site and apply to secondary glance

A3B5A3B5

A3B5

A3B5

Page 41: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation41

Automation Example: Nova

1. Provision VM at primary (possibly with Heat)

Primary

Nova

Flavorsfoo

barbaz

VM1:

foo

VM2:

baz

Secondary

Nova

Flavors

baz

Page 42: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation42

Automation Example: Nova

1. Provision VM at primary (possibly with Heat)

2. Extract metadata and dependencies, e.g., flavors, from

primary and replicate to secondary

Primary

Nova

Flavorsfoo

barbaz

VM1:

foo

VM2:

baz

Secondary

Nova

Flavors

baz

VMs

VM1:foo; VM2:bazDependencies

Flavors: foo, baz

Page 43: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation43

Automation Example: Nova

1. Provision VM at primary (possibly with Heat)

2. Extract metadata and dependencies, e.g., flavors, from

primary and replicate to secondary 3. Create dependencies and heat template at secondary

Primary

Nova

Flavorsfoo

barbaz

VM1:

foo

VM2:

baz

Secondary

Nova

Flavorsfoo

baz

VMs

VM1:foo; VM2:bazDependencies

Flavors: foo, baz

Template

Provision

VM1 fooVM2 baz

Page 44: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation44

Automation Example: Nova

1. Provision VM at primary (possibly with Heat)

2. Extract metadata and dependencies, e.g., flavors, from

primary and replicate to secondary 3. Create dependencies and heat template at secondary

4. Use template to deploy on failure

Primary

Nova

Flavorsfoo

barbaz

VM1:

foo

VM2:

baz

Secondary

Nova

Flavorsfoo

baz

VM1:

foo

VM2:

baz

VMs

VM1:foo; VM2:bazDependencies

Flavors: foo, baz

Template

Provision

VM1 fooVM2 baz

Page 45: Surviving the Worst · 2019-02-26 · 1 © 2013 IBM Corporation Surviving the Worst: A Vision for OpenStack Disaster Recovery Presenter: Michael Factor factor@il.ibm.com IBM Research

© 2013 IBM Corporation45

Outline and Presentation Objective

Basic Disaster Recovery (DR) concepts

Workload Example

Vision for OpenStack DR

Objective: Motivate the requirements for DR for OpenStack and encourage involvement in on-going efforts to enable OpenStack DR.

We are just scratching the surface and encourage involvement.

• See https://wiki.openstack.org/wiki/DisasterRecovery

• Cinder design summit session on volume continuous replication today at 2:40 PM.

• Unconference session tomorrow at 9:50am